KEYNOTE: Image processing, AI, and autonomous systems
summary
Overview of image-processing tasks with concrete solutions and failure modes: ship detection in harbors (example contrasts MS COCO (80k) pretraining with 1,600 infrared images fine‑tuning), search‑and‑rescue robots and Spot-style manipulators that learn to find handles using Segment Anything and CLIP, and pipeline design emphasizing edge inference, mission/runtime updates, and validation. Claims: foundation models (GPT/CLIP) can unify perception and planning but introduce bias, limited reproducibility, and open‑world robustness gaps that block immediate operational use.
AI-written summary · about
chapters (9)
- 0:00Image processing AI intro
- 1:30Situational awareness from images
- 7:28Autonomous systems limits
- 14:27Open-world challenges
- 20:28AI evolution to generative models
- 25:10Military vs civil AI challenges
- 31:34Generative AI in practice
- 40:43Future of generative AI
- 41:05Closing & Q&A
related talks
24mAI in maritime - Autonomous ships
36mAI-Powered Real-Time Quality Control
40mKeynote: Open-source Multimodal AI in the Wild
01h 00mOpening Keynote
04mAI and the Robotics Revolution
46mML Use Cases at Think Big Analytics
19mAI Trends in Intelligent Transport Systems
41mDeep Learning to Solve Challenging Problems